valbuena et al 2008 aee

8/13/2019 Valbuena Et Al 2008 AEE

http://slidepdf.com/reader/full/valbuena-et-al-2008-aee 1/10

A method to define a typology for agent-based analysis in regionalland-use research

Diego Valbuena *, Peter H. Verburg, Arnold K. Bregt

Department of Environmental Sciences, Wageningen University, PO Box 47, 6700 Wageningen, The Netherlands

1. Introduction

Land-use/cover change (LUCC) is a complex process caused by

the interaction between natural and social systems at different

temporal and spatial scales (Lambin and Geist, 2001; Rindfuss

et al., 2004). In rural areas, LUCC is related to the dynamics of the

agricultural sector in general, and those of the farming systems in

particular. However, each farming system is different (Kobrich

et al., 2003), reflecting the heterogeneity of human’s behaviour and

decisions. This heterogeneity is not only influenced by the

complexity of the human behaviour itself (Simon, 1955; Rokeach,

1968; Ajzen, 1991; Arthur, 1994), but also by both internal and

external factors such as land-holders’ experience, family structure,

economic and technical resources, and the socio-economic context

where these decisions occur (Fig. 1) (Gasson, 1973; Ilbery, 1978;Evans and Ilbery, 1989; Willock et al., 1999b; Knowler and

Bradshaw, 2007).

An approach to analyse and simulate human decisions in LUCC

is theuse of multi-agentsystems (MAS) (Parkeret al., 2003; Brown,

2005; Matthews et al., 2007; Robinson et al., 2007). MAS are

modelling tools in which decision-making entities are represented

by agents, while the environment is defined by spatial data. This

conceptualization of the land-use system explicitly includes

interactions between agents and between agents and the

environment. Agents in the model can represent individuals or

groups of people, institutions, etc. Agents can be designed with

different characteristics: they can be heterogeneous (e.g. socio-

cultural background, economic situation, perception of the land-

scape, age, behaviour), autonomous (they take their own land-use

decisions based on rational and non-rational rules) and dynamic

(they can learn and adapt to different situations) (Ferrand, 1996;

Bonabeau, 2002; Parker et al., 2003; Sawyer, 2003; Crawford et al.,

2005). By using autonomous and heterogeneous agents, MAS

explicitly deal with the diversity of land-use decisions. In this way,

MAS cope with the limitation of most of the land-use models

implemented at a regional scale (Balmann, 2000; van derVeen and

Otter, 2001; Bonabeau, 2002; Sawyer, 2003; Evans and Kelley,2004; Verburg, 2006), which often use a single response function

throughout the study area, assuming that human decision-making

is a homogeneous process (e.g. Fohrer et al., 2002; Soares-Filho

et al., 2002; Verburg et al., 2002; Luijten, 2003).

Agents can be defined and parameterized in many different

ways, depending on the objectives of the MAS itself ( Janssen and

Ostrom, 2006; Robinson et al., 2007). For example, agents can

represent broad groups of stakeholders such as landowners,

government or environmentalists (e.g. Ligtenberg et al., 2004;

Monticino et al., 2007); socio-economic units such as households

(e.g. Evans and Kelley, 2004; Matthews, 2006); or organizational

Agriculture, Ecosystems and Environment 128 (2008) 27–36

A R T I C L E I N F O

Article history:Received 12 November 2007

Received in revised form 25 April 2008

Accepted 28 April 2008

Available online 13 June 2008

Keywords:

Land use/cover change

Multi-agent systems

Land-use decisions and strategies

Agent definition and parameterization

Agent typology

A B S T R A C T

Land use/cover change (LUCC) is often the cumulative result of individual farmer’s decisions. Tounderstand and simulate LUCC as the result of local decisions, multi-agent systems models (MAS) have

become a popular technique. However, the definition of agents is not often based on real data, ignoring

the inherent diversity of farmers and farm characteristics in rural landscapes. The aim of this paper is to

describe an empirical method thatdefines an agent typology and allocates agents into the different agent

types for an entire region. This method is illustrated with a case study in the Netherlands, where

processes of farm expansion and diversification of farm practices take place. Five different agent types

were defined and parameterized in terms of views, farm characteristics and location. Despite its

simplicity, this empirical method captures several relations between farmers’ views, farm characteristics

and land-use decisions and strategies.This approach is a stepforward in multi-agent systems of landuse/

cover change (MAS/LUCC) to include the diversity of land-use decisions and strategies in regional studies

by empirically defining, parameterizing and allocating different agent types.

2008 Elsevier B.V. All rights reserved.

* Corresponding author. Tel.: +31 317482069; fax: +31 317419000.

E-mail address: [email protected] (D. Valbuena).

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

j o u r n a l h o m e p a g e : w w w . e l s ev i e r . c o m / l o c at e / a g e e

0167-8809/$ – see front matter 2008 Elsevier B.V. All rights reserved.

doi:10.1016/j.agee.2008.04.015

mailto:[email protected]

http://www.sciencedirect.com/science/journal/01678809

http://dx.doi.org/10.1016/j.agee.2008.04.015

http://dx.doi.org/10.1016/j.agee.2008.04.015

http://www.sciencedirect.com/science/journal/01678809

mailto:[email protected]



units such as farms (e.g. Balmann, 2000; Happe et al., 2006). The

decision-making process of agents needs to be specifically

parameterized by decision rules. These rules can be defined

based on either artificial or empirical data. Agent parameteriza-

tion with artificial data is the more common approach in MAS

(Berger and Schreinemachers, 2006). Although improving the

theoretical insight into LUCC dynamics, the validity of such

artificial agent definition is often difficult to assess (Gimblett,

2002; Parker et al., 2003). On the other hand, agent parameter-

izationwith empirical datacan facilitate theunderstanding of real

LUCC processes. Still, most studies using empirical data to define

and parameterize agents rely on intensive data gathering (e.g.

Bousquet et al., 2001; Huigen, 2004; Castella et al., 2005; Jepsenet al., 2006).

At a regional scale, intensive data gathering is limited by the

large number of agents. Despite the heterogeneity of farming

systems at such spatial scale, general farming strategies can be

distinguished, what Bowler (1992) define as ‘‘paths of develop-

ment’’ (see also Wilson, 2007). These general pathways are a

simplificationof how farming can developin a certainarea(Meert

et al., 2005). A relevant approach to analyse the heterogeneity in

decision-making of farmers is to formulate typologies (McKinney,

1950; Jollivet, 1965; Escobar and Berdegue, 1990; Perrot and

Landais, 1993b; van der Ploeg, 1994). A typology is a tool to

simplify the diversity of farmers and farming strategies. This

means that a typology is an artificial way to define different

groups based on specific criteria in order to organize and analysereality (McKinney, 1950; Jollivet, 1965). Thecriteria to construct a

typology, as well as to evaluate it, primarily depend on the

objectives of its implementation (Escobar and Berdegue, 1990).

Different kinds of typologies for agents in rural areas can be

distinguished based on their aim. For example, some typologies

intend to understand the whole farming process, which include

the most relevant farm(er) characteristics (e.g. Escobar and

Berdegue, 1990; Perrot and Landais, 1993a; van der Ploeg, 1994).

Other typologies aim to analyse the underlying reasons of certain

farmers’ decisions (e.g. Morris and Potter, 1995; Fish et al., 2003;

Guillaumin et al., 2004). Finally, other typologies aim to explain

the different production strategies that farmers developed or

might develop (e.g. Ondersteijn et al., 2003; de Lauwere, 2005;

Vanclay et al., 2006; Van Doorn and Bakker, 2007). Nevertheless,most of the current typologies do not account for the spatial

linkage in which land-use decisions are embedded (Landais,

1998).

The aim of this study is to describe an empirical method to

define and parameterize differentagent types for use in MAS/LUCC

at a regional scale. This method first defines an agent typology and

subsequently distributes spatially the defined agent types in the

entire region. After describing the general method, this paper

illustrates the method with a case study in the East of the

Netherlands where processes of farm diversification and farm

expansion are taking place. Finally, this paper discusses the

advantages and the limitation of the proposed method, including

its potential applications in MAS/LUCC research and policy-making

processes.

2. General method

To include thediversity of land-use decisions and strategies at a

regional scale in a MAS/LUCC, we need first to simplify the diversity

between all individual agents by formulating an agent typology

(typology formulation, Fig. 2), andthereafter,we need to distributespatially the defined agent types in the region (agent type spatial

distribution, Fig. 2).

2.1. Typology formulation

In the first part of the method, the diversity of land-use

decisions and strategies is simplified by defining and characteriz-

ing different agent types in the region based on four steps ( Fig. 2).

2.1.1. Step 1. Agent modelling objectives

Before simplifying the diversity of farmers’ decisions and

strategies, we need to define clearly the aim of such simplification.

Inthe case ofMAS at a regional scale, weneed toestablish what the

agent modelling objectives are: do we want to understand andsimplify the whole LUCC dynamics of an area? Are we interested

only in simulating a particular process? Or are we interested in the

reasons to follow a particular land-use management strategy?

2.1.2. Step 2. Criteria selection

The definition of the agent modelling objectives facilitates the

selection of different criteria that are used to define and

parameterize the agent typology. The selection of these criteria

not only reflects the objective of the research and the scientific

approach (e.g. Berdegue and Escobar, 1990), but also defines how

the diversity of land-use decisions and strategies is simplified and

included in MAS/LUCC. Still, such criteria should not rely only on

quantitative analyses, but they also need to have a meaning to the

social reality of LUCC (Perrot and Landais, 1993b; Landais, 1998).Further, these criteria generally need to describe the views and

perceptions of the agent as well as their socio-economic situation

and context. Related to views and perceptions, Burton and Wilson

(2006) argue that farmers can have different identities (e.g.

agricultural producer and diversifier). In this way, a typology can

be defined by using a combination of identities. Data to define

these identities are often not available for all agents in a large

region because restrictions on data accessibility or the data do not

even exist. Therefore, such data can be gatheredby using a detailed

sample survey.

2.1.3. Step 3. Agent type definition

The definitionof different agent types primarily depends on the

selected criteria. Different methods to construct typologies have

Fig. 1. Influences on agent’s land-use decisions.

Fig. 2. Diagram of the method.

D. Valbuena et al. / Agriculture, Ecosystems and Environment 128 (2008) 27–36 28



been described (e.g. Escobar and Berdegue, 1990; Landais, 1998;

van der Ploeg, 2003). For example, a typology can be constructed

using qualitative (e.g. Guillaumin et al., 2004; Schmitzberger et al.,

2005) or quantitative analyses (e.g. Wilsonand Hart, 2000; Kobrich

et al., 2003; Kristensen, 2003; de Lauwere, 2005); attitudinal (e.g.

Fairweather and Keating, 1994; Morris and Potter, 1995) or socio-

economic variables (e.g. Commandeur, 2005; Schmitzberger et al.,

2005); and scientist knowledge (e.g. Perrot and Landais, 1993a;

van der Ploeg, 1994) or participatory processes (e.g. Girard, 2006).

The choice of a particular analysis depends on the selected criteria

and the available data.

2.1.4. Step 4. Agent type parameterization

In this step, agent types are described in terms of views and

strategies relevant to the MAS/LUCC. This is a step to understand

the differences between agent types based not only on views

and decisions, but also their socio-economic situation and

context. To assess whether these differences are significant,

statistical analyses are used. Most of the attitudinal and

socio-economic variables used to parameterize agent types

are included in the detailed sample survey. Finally, some of

these variables can be also used as common variables (see

below).

2.2. Agent type spatial distribution

Inthe secondpartof the method, all agents inthe study area are

classified as one of the defined agent types. Further, the method is

validated and verified.

2.2.1. Step 5. Agent type allocation

Although full census data of the entire population of a region

are often available, the level of detail and/or accessibility of such

data areoften limited. Therefore, these data arenot used to define

the different agent types. The use of common variables between a

detailed sample survey, full census data and additional socio-

economic and biophysical dataallows us to classify all agents intoone of the defined agent types (Fig. 2). Although spatial variables

canbe usedto define andparameterize thedifferent agent types, it

is in this step that agent types are explicitly linked to the

landscape. Finally, depending on the available data for the entire

study region, different quantitative analyses techniques can be

used to classify agents, such as regression, multivariate and

probabilistic analyses.

2.2.2. Step 6. Validation/verification

Thedefinition andallocation of the different agent types (step 3

and 5, Fig. 2) may result in some degree of overlap between agent

types and consequently uncertainty about the definition of the

variables due to both the complexity of land-use decisions and the

limited data availability. For this reason, it is necessary to validatethese steps and determine the level of uncertainty of the results.

Based on this validation/verification process, early steps of the

method might need to be modified. Participatory approaches could

be used to validate an agent typology (e.g. Girard, 2006). However,

agent types arean abstract representationof reality whichmakesit

difficult for people to recognize themselves in the agent types as

defined and therefore hampers such a validation.

3. Case study

In this part, theproposed method is illustrated with a case study

in the eastern part of the Netherlands, where processes of

diversification of farm practices and farm expansion are reshaping

the landscape structure.

3.1. Study area

The study area is located in the eastern part of the Netherlands

and covers an area of approximately 60.650 ha (Fig. 3). By 2005,

there were around 2300 agricultural holdings; about 66% of them

were livestock farms (Farm Accountancy Data). Part of this area

represents a cultural–historic landscape where small-scale agri-

culture and nature areas are closely related providing a particularcultural, recreational, tourist, ecological and economic value to the

region (Provinciale Staten van Gelderland, 2005b). The spatial

structure of the landscape has been the result of the interaction

between biophysical (e.g. soil characteristics and water avail-

ability) and socio-economic factors and processes (e.g. land tenure,

accessibility and labour demand) (Benvenuti, 1961; Wildenbeest,

1989; Mastboom, 1996).

In the last decades, social changes such as the increasing

environmental awareness and the growing demand for recreation

and tourist areas, together with legislative changes such as the

establishment of milk quotas, restrictions on manure applications,

and compensatory payments for nature and landscape conserva-

tion have influenced the rural dynamics not only in the study area

(Provinciale Staten van Gelderland, 2005a), but also in most of therural areas in the Netherlands (van Horne and Prins, 2002;

Graveland et al., 2004; Oerlemans et al., 2004; Berkhout and

Bruchem, 2005).

3.2. Input data

This study is based on a detailed sample survey carried out

during winter 2004 ( Jongeneel et al., 2005), full census data of the

entire population in the study region (farmer accountancy data)

and additional socio-economic and biophysical spatial data. The

detailed sample survey included 333 farmers and it was originally

carried out to explore the factors that determine the diversification

of farm practices including farmer’s views (positive, neutral and

negative) and structural variables such as the existence or not of a

Fig. 3. Study area.




successor, production scale, degree specialization of the farm and

past land-use changes ( Jongeneel et al., 2005). The full census data

include only a limited number of structural variables such as

agribusiness type, production scale, cultivated area, head farmer’s

age and labour, and no attitudinal variables. Finally, additional

spatial data include: groundwater table, altitude, proportion of

nature and historical areas, farm density, production scale and

cultivated area.

To simplify the parameterization and allocation of the different

agent types, some variables were grouped. In particular, agribusi-

ness type and category of production scale were defined based on

CBS (Statistics Netherlands) terminology. Four main agribusiness

types from the Dutch version of the Community Typology were

distinguished for the study area: arable, livestock, intensive

livestock and other farms. Also, based on the production scale,

farms were divided into four main categories: hobby (3–20 dsu),

small (20–50 dsu), medium (50–100 dsu) and large farms

(>100 dsu). Dsu or Dutch size units represent the economic size

of a farm including the amount and use of the land; in 2005, a dsu

was equal to 1400 euros.


3.3.1. Step 1. Agent modelling objectives

The agent modelling objective is to understand the different

strategies that farmers have followed and/or might follow in terms

of farm diversification practices and farm expansion.

3.3.2. Step 2. Criteria selection

Because these land-use decisions often depend on individual

decisions, the criteria to define this typology should include

farmers’ views or willingness (Siebert et al., 2006). In particular,

views on expansion of production scale as a future alternative,

diversification of farm practices as an additional income and

participation in compensation schemes for nature and landscape

conservation practices were selected. At the same time, socio-

economic variables can implicitly represent different farmingstrategies. This is partly related to the farmer’s ability to carry out

certain action (Siebert et al., 2006). Specifically, farming is an

important source of income for some farmers or it is simply a

hobby for others. This distinction has an influence in decisions

concerning the diversification of farm practices and the expansion

of their holdings (e.g. Primdahl, 1999; Kristensen, 2003; Schmitz-

berger et al., 2005). Therefore, both socio-economic and attitudinal

variables are used as criteria to define the different agent types.

3.3.3. Step 3. Agent type definition

Based on an analysis of the diversity in views and socio-

economic conditions, a ‘‘classification tree’’ was chosen as most

appropriate method to construct the typology in this case study.

This tree is based on Boolean statements defined in the criteriadescribed above: production scale larger than 20 dsu, view on

expansion of the production scale and on participating in

compensation schemes. The results of this classification tree were

supported by explorative analyses, including regression, factor and

cluster analysis.

3.3.4. Step 4. Agent type parameterization

After agent types were defined, they were parameterized based

on additional variables suchas age, education, cultivated area, type

of agribusiness, successor, past land-use decisions, membership of

different organizations, knowledge about different agricultural

projects, etc. To define whether the differences of the metric

variables between agent types were significant, analysis of

variance (ANOVA) was used.


3.4.1. Step 5. Agent type allocation

In this case study, full census data were available for the entire

agent population. This means thatsocio-economic data of eachreal

agent and its exact location were available. Therefore, a

classification of these agents into agent types could be made

using the common variables between the detailed survey data and

census data. The common variables included: age, agribusiness

type, production scale and landscape structure.

The analysis of the landscape structure was carried out at

postcode level (average area 550 ha), at which the location of the

farms of the detailed survey was known. To calculate the relation

between these spatial variables, Pearson correlation analyses were

carried out. Finally, to assess whether the landscape structure was

significantly related to the agent typology or farming decisions

(non-parametric variables), Kruskal–Wallis tests were carried out

(for detailed information on these analyses see Legendre and

Legendre, 1998; Lesschen et al., 2005).

To classify the different agent types a classification tree was

calculated using the CRT growing method (SPSS 15.0). This

approach is similar to the one discussed by Speybroeck et al.

(2004), who selected as splitting criteria the ‘Gini method’ becauseit performs the best. To avoid overfitting the model, the size of

parent and child nodes was limited to 20 and 4 respectively. The

definition of these sizes was related to both the size of the detailed

sample survey and the validation procedure (see below). The

advantages of using classification trees are that it allows

combining metric and non-metric variables, and including non-

linear relationships.

3.4.2. Step 6. Validation/verification

To assess the uncertainty of the classification process, a cross-

validation method for the classification tree was implemented

(SPSS 15.0). With this method, the dataset was divided into 25 sub-

samples. Each sub-sample was classifiedbasedon the results of the

classification tree of the other 24 samples. The proportion of casesthat were correctly classified was calculated for the 25 runs.

Further, the classification tree assigns to each agent the probability

to belong to each agent type, selecting the type that hasthe highest

probability. To see whether this probability was the same

throughout the landscape, a map of the mean highest probability

was calculated (ArcMap 9.2).

4. Results

In the study area, there is a tendency towards the reduction of

the production scale. While around 7% of farmers belonged in 2005

to a higher category of production scale than in 2001, almost 30%

belonged in 2005 to a lower category. Furthermore, according to

the CBS data, between 2001 and 2005 there was a decrease of 12%of farm holdings in the entire province (around 1950 holdings).


As it was mentioned before, the agent modelling objective in

this case study is to understand and simplify the different

strategies that farmers have followed and/or might follow in

terms of farm diversificationpractices and farm expansion (step1).

Two different criteria to define agent types were selected:

production scale and farmers’ views (step 2). This first distinction

reflects to some extent the role of agriculture for a farmer (life-

style vs. life-style and production). It also distinguishes farmers

with very small farms who might have a relative less relevant role

in landscape dynamics than those with bigger farms. This is




partially reflected by the fact that most farmers with hobby farms

(<20 dsu) had similar views: they were not willing to expand their

production scale (84%) and they did not have a negative view on

compensation schemes for nature and landscape conservation

practices (95%).

The second distinction reflects the current and potential

probability that a farmer diversifies farm practices and/or expands

the productionscale of his/her holding. In fact, around 80%of those

who think that a future alternative is to increase the production

scale have increased the area of their farm and the milk quotabetween 2001 and 2005. On the other hand, around 17% of those

who have a negative view on participating on compensation

schemes have diversified farm practices, compared to 32% of those

with neutral view and 51% of those with a positive view. However,

there were no differences in terms of farm diversification between

those farmers with a negative and those with a neutral view on

participating on such schemes.

Based on these two criteria and the fact that most farmers did

not disagree on seeing diversification of farm practices as a means

to obtain more income, five different agent types were defined

(step 3): hobby farms (H), non-expansionist conventional (NC);

non-expansionist diversifiers (ND); expansionist conventional

(EC); and expansionist diversifiers (ED) (Fig. 4).

These agent types can be parameterized based on several socio-

economic characteristics (step 4) (Tables 1–4).

- Hobby farms: agents with different ages and with different

agribusiness types who own very small farms and who normally

have an off-farm job. They are unlikely to expand the production

scale of their farms. In fact, a relatively high proportion of them

have decreased the size of their farm and the milk quota.

Moreover, the probability that they would stop farming is

relatively high, whereas the probability that they participate in

different associations or that they know about different

agricultural projects is relatively low. Also, they are unlikely to

participate in compensation schemes. However, the probability

that they implement other farm practices such as tourism and

recreation is relatively high.- Non-expansionist conventional: in general, agents who are not

young and who do not own arable farms. They own most of their

land and they have different categories of production scale. Still,

they are unlikely to expand their production scale. The

probability that they stop farming is high and that they

Fig. 4. Agent type definition.

Table 1

Agent type parameterization: production strategies

Agent type Stop farming Increase

production

Decrease

production

Diversify farm

practices

Compensation

schemes

Tourism and

recreation

Hobby farm + + +

Non-expansionist conventional + + +/ +/ +

Non-expa nsio nist dive rsi fier +

+ + + +Expansionist conventional +

Expansionist diversifier + + + +

Table 2

Agent type parameterization: age and farm characteristics (mean and standard deviation)

Agent type No. Age* Agricultural

income (prop)

Production

scale (dsu)

Cultivated

area (ha)

Labour

(h/week)

Lab. Family

(h/week)

Owned

land (prop)

H 43 49 (9) 0.58 (0.30) 10.5 (5.1) 15.1 (14.7) 25.6 (15.3) 12.6 (18.6) 0.85 (0.26)

NC 44 52 (11) 0.81 (0.25) 96.4 (61.9) 24.7 (18.0) 40.0 (9.5) 28.5 (24.2) 0.76 (0.32)

ND 34 50 (9) 0.71 (0.34) 77.57 (61.3) 24.0 (14.9) 39.7 (9.0) 17.5 (16.3) 0.70 (0.34)

EC 115 48 (10) 0.71 (0.28) 125.0 (65.2) 36.2 (24.0) 42.3 (6.6) 32.8 (21.6) 0.66 (0.34)

ED 48 48 (12) 0.77 (0.30) 143.2 (78.8) 39.8 (20.6) 41.9 (7.8) 40.1 (24.0) 0.68 (0.26)

* Excluding age, all the variables were significantly different between agent types (ANOVA, p < 0.05).

Table 3

Agenttype parameterization: education, off-farm job, farmdiversification and expansion indicators(numbersindicate percentage of farmsbelonging to a class, participating

in diversification activities or showing respectively an increase or decrease in size)

Agent type No. Education: low level Off-farm job Divers. farm Nature and landscape Tourism-recreationa Farm sizea Milk quotaa

Increase Decrease Increase Decrease

H 43 36.8 56.1 22.0 19.5 7.3 7.3 12.2 14.6 22.0

NC 44 35.7 20.5 34.1 31.8 4.5 34.1 11.4 31.8 9.1

ND 34 26.4 23.5 52.9 47.1 5.9 41.2 17.6 26.5 14.7

EC 115 18.7 19.5 21.2 20.4 0.9 66.4 1.8 76.1 0.9

ED 48 10.6 23.4 46.8 46.8 4.3 70.2 0 68.1 4 .3

a

In the last five years.




participate in different associations or that they know about

different agricultural projects is relatively low. Although they do

not have a positive view towards compensation schemes, the

proportion of diversification of farm practices is relatively higher

than that of hobby farms.

- Non-expansionist diversifiers: agents who are neither young nor

old, mainly with livestock and intensive livestock farms. They

own small and medium farms, and they normally own most of

their land. They are unlikely to expand their production scale,

and many of them have decreased their farm size and their milk

quota. The probability that they stop farming, that they

participate in different associations, that they know aboutdifferent agricultural projects and that they diversify farm

practices is relatively high.

- Expansionist conventional: agents with different ages, and with

livestock and intensive livestock farms. They own medium and

large farms and only half of them own most of their land. The

probability that they increase the production scale of their farms

is high, and that they decrease the production scale or that they

stop farming is low. Their participation in agricultural organiza-

tions is high, excluding those for nature and landscape

conservation. The probability that they diversify farm practices

is low.

- Expansionist diversifiers: agents who are relatively young with

relatively high level of education. They own medium and large

farms with different agribusiness types. The probability that they

increase the production scale of their farms is high and that they

stop farming is low. Their participation in agricultural organiza-

tions is the highest. The probability that they diversify farm

practices is also high, mainly their participation in compensation

schemes.


The common variables include age, agribusiness type, produc-

tion scale and landscape structure. In terms of age, around 40% of

the agents of type ED are less than 40 years old and of type EC are

between 40 and 50 years old. In terms of production scale, almost

42% of agents of type ND own small farms, while more than 60% of

type EC and ED own large farms. In terms of agribusiness type,

Table 4

Agent type parameterization: views, membership and acquaintance with projects (percentages)

Agent type No. Stop farming Agriculture

organization

Local parties: improve

agricultural sector

Agriculture association:

nature and landscape management

Acquaintance of different

agriculture projects

H 43 34.1 46.3 7.3 24.4 65.9

NC 44 36.4 77.3 6.8 13.6 61.4

ND 34 32.4 76.5 26.5 47.1 88.2

EC 115 6.2 89.4 18.6 16.8 84.1

ED 48 4.3 91.5 34.0 53.2 85.1

Fig. 5. Densities of each agent type. (A) H, (B) NC, (C) ND, (D) EC, and (E) ED.




most of the arable farms belong to agents of type H, while about

82% of the agents of type EC own a livestock farm and relatively

more intensive farms belong to agents of type NC and ND. Finally,

about half of the agents of type H are located in postcode areas

with a proportion of nature areasbetween 5% and 20%, whileonly

29% of the agents of type ED are located in those areas. Also,

around 50% of agents of type H and 41% of those of type ND are

located in areas with small-scale agriculture—postcode areas

where parcels smaller than 2 ha represent more than 20% of the

total area—while only 32% of those of type ED are also located in

those areas.

The landscape structure is significantly related to the diversi-

fication of farm practices. Agents who have diversified farm

practices tend to be located in postcode areas higher in altitude

( Z = 1.93 p< 0.1) and with more nature ( Z = 2.02 p < 0.05)

where the average parcel size is relatively smaller ( Z = 3.45

p< 0.01) than those who have not diversified. At the same time,

some of these variables are correlated. Nature tends to be located

in higher areas (r = 0.54, p < 0.01) where the groundwater table is

lower (r = 0.25, p< 0.05). Also part of the nature areas have a

historical value in the study area (r = 0.36, p < 0.01). On the other

hand, postcode areas with a high nature density tend to have a

higher farm density (r = 0.55, p< 0.01), more hobby farms(r = 0.27, p< 0.05), less intensive livestock farms (r = 0.26,

P < 0.05) and more farms with tourism and recreational facilities

(r = 0.30, p < 0.01).

The classification process shows that each type is not equally

distributed throughout the region (step 5). This distribution is

partly caused by the spatial structure of the landscape: small-scale

agriculture towards the South-West and large-scale agriculture in

some parts in centre of the study area (Fig. 5). For example, the

density of hobby farms is higher near urban and nature areas

(Fig. 5A), where small-scale agriculture often takes place. In

contrast, the density of EC and ED is higher in areas where large-

scale agriculture occurs (Fig. 5D and E).

Excluding hobby farms, the cross-validation of the classification

tree (step 6) shows that around 50% of the agents of the detailedsurvey were correctly classified using this approach. This

percentage is higher than if the entire allocation/classification

process is carried out randomly (25%). This degree of uncertainty

was mainly caused by the overlap between different agent types.

This overlap is related to the distribution of the common variables

within each agent type. For example, the overlap between agent

types EC and ED was relatively high, making the distinction

between these two types less clear than in other cases. This

uncertainty is reflected differently among the different agent

types. As it was mentioned before, an agent is assigned to the type

in which he obtained the highest probability. Thus, the highest

probability that an agent belong to a particular type varied among

types. In particular, about 66% of the agents of EC had a probability

to belong to this agent type lower than 60%, while about 88% of those of ED had a probability higher than 80%. Because the agent

distribution is not homogenous throughout the region, then the

uncertainty of the classification process throughout the region is

also heterogeneous (Fig. 6).

When looking at therelative amount of agentsand themeans of

age, production scale and cultivated area of each agent type in the

results of the construction of the agent typology (Table 3) and in

the results of the classification procedure (Table 5), several

differences can be distinguished. The mismatch between these

two set of results indicates that the detailed survey does not

completely represent the entire population. In fact, small-scale

dairy farms (mainly hobby farms) are underrepresented, whereas

large-scale dairy farms and young managers are overrepresented

in the survey.

5. Discussion

The construction of an agent typology is the first step to include

the diversity of farmers and/orfarming strategies in LUCC research

at a regional scale, in particular in MAS. The linkage between this

typology to the landscape allows us to combine different kind of

data at different scales. These different kinds of data can be also

seen as individual empirical methods to define agent-based

models, including survey samples, census and GIS data ( Janssen

and Ostrom, 2006; Robinsonet al., 2007). Other empirical methods,

such as interviews, experiments and observations could be as wellincluded to define an agent typology. However, it is the

combination of these different input data that allows allocating

empirically the different agent types through an entire region,

where there is normally a lack of detailed data. Consequently, this

approach makes it possible to simplify and include the diversity of

farming systems and individual decisions in MAS/LUCC at a

regional scale.

The method described in this paper has some limitations. The

different agent types distinguished in this study overlap in terms of

farm(er) characteristics and decisions. Although characteristics

such as age and agribusiness type might be important indicators of

farmers’ views, perceptions and decisions, these are interacting

with so many other variables that they are unlikely to discriminate

perfectly agent types. In other words, these characteristics make it

Fig. 6. Highest probabilities to belong to a particular agent type.

Table 5

Average of the means and standard deviation allocated agent types

Agent type No. Age Production

scale (dsu)

Cultivated

area (ha)

H 735 59 (12) 9.1 (4.5) 6.8 (4.5)

NC 403 54 (12) 69.7 (41.8) 23.5 (14.2)

ND 296 55 (11) 35.3 (10.6) 14.5 (9.6)EC 522 54 (12) 113.0 (60.3) 38.0 (17.7)

ED 136 52 (11) 126.8 (62.8) 28.8 (21.9)




possible to distinguish general farming strategies at a regional

scale, but describing the differences of socio-economic character-

istics between these strategies is difficult. This is in line with the

statement of Vanclay et al. (2006) that types always overlap

because the vast diversity of farm(er)s.

Views and attitudes of farmers are assumed to have a major

influence on land-use decisions and strategies (Fairweather and

Keating, 1994; Willock et al., 1999a; Busck, 2002; Sharpley and

Vass, 2006). However, the data showed exceptions. For example,

althoughsome agents were not supposed to diversifytheirfarming

practices based on their views, they did it in reality. This supports

the claim of Burton and Wilson (2006) that farmers have different

identities, depending on the circumstances when they take their

decisions. Therefore, typologies do not represent reality perfectly,

however, they are relevant tools whose use can facilitate our

understanding of land-use dynamics (McKinney, 1950; Jollivet,

1965; Perrot and Landais, 1993a; Fish et al., 2003; Vanclay et al.,

2006).

The method presentedin this paper depends on the amount and

quality of the data. Thus, it requires a set of common variables

through whichdifferent kinds of data canbe linked. In addition,the

information that farmers supply to registration systems might not

always match with the real data (van der Ploeg, 2003). Related tothis, a detailed sample survey that tries to cover a whole region

does not necessarily represent the entire population, leading to a

mismatch between the detailed sample survey and the statistical

data used for allocation/classification.

Nevertheless, the agent typology formulated for the study area

included many of the interactions that have been described in

literature between farmers’ views, farm(er) characteristics and

currentrural processes suchas diversificationof farm practicesand

farm expansion (e.g. Kelly and Ilbery, 1995; Austin et al., 1996;

Willock et al., 1999a; Knickel and Renting, 2000; Wilson and Hart,

2000; Ondersteijn et al., 2003; Knowler and Bradshaw, 2007;

Jongeneel et al., 2008). Moreover, most of agent types defined in

this researchare related to one or to the combination of some of the

identities that currently define European farmers (Burton andWilson, 2006). All this suggests that the agent typology presented

in this paper has been able to capture the diversity of land-use

decisions and strategies occurring in rural landscapes.

The analyses of the case study also confirm previous studies

that showed that spatial location of a farm can influence land-use

decisions and therefore farming strategies (Bryant and Johnston,

1992; Luttik and van der Ploeg, 2004; Jongeneel et al., 2008). These

results also showed that the spatial distribution of such decisions

and strategies are not randomly distributed throughout the

landscape, which is in accordance with the findings of Kobrich

et al. (2003). By defining agent typology and linking it with the

landscape, the approach proposed in this paper allows us to link

agents’ decisions to their environment, which is one of the key

aspects of LUCC/MAS (Balmann, 2000; Parker and Berger, 2002;Evans and Kelley, 2004). The empirical definition of an agent

typology differs with theagentdefinition of most MAS that assume

more ‘‘stereotype’’ agents that assume very different character-

istics and decisions. In fact, the results suggest that defining

‘‘stereotype’’ agents may oversimplify the real diversity in

decision-making, restricting their implementation to more theo-

retical problems.

According to Geertman and Stillwell (2003), Uran and Janssen

(2003) and McIntosh et al. (2007), the lack of clarity and flexibility

represents a barrier to implement land-use models and decisions

support-systems in planning and policy-making processes.

Because of its clarity and flexibility, the approach described in

this paper moves thus towards a better adoption of MAS/LUCC in

such processes. The flexibility of this approach makes it also

suitable to be used in different cases studies, where farming

systems and rural dynamics might be different.

In conclusion, the method described in this paper is a simple

and straightforward approach that combines different empirical

methods to built agents-based models. This combination, together

with the spatial definition of different agent types, allows us to

parameterize, allocate and validate different agent types in an

entire region. This empirical method is a step forward to include

the diversity of land-use decisions and strategies in MAS/LUCC at a

regional scale, as well as the spatial dimension where these

decisions take place.

Acknowledgements

The authors gratefully acknowledge Nico Polman and Roel

Jongeneel for allowing us to use their dataset; the LEI, in particular

Tom Kuhlman for the access to the Farm Accountancy Data; to the

Province of Gelderland for proving us with the spatial data; and to

the three anonymous referees who provide us with relevant

comments to improve this paper.

References

Ajzen, I.,1991. The theory of planned behavior. Organizational Behavior and HumanDecision Processes 50, 179–211.

Arthur, W.B., 1994. Inductive reasoning and bounded rationality. American Eco-nomic Review 84, 406–411.

Austin, E.J., Deary, I.J., Gibson, G.J., McGregor, M.J., Dent, J.B., 1996. Attitudes andvaluesof Scottish farmers: ‘‘Yeoman’’and ‘‘entrepreneur’’ as factors, not distincttypes. Rural Sociology 61, 464–474.

Balmann, A., 2000. Modeling land use with multi-agent systems. Perspectives forthe analysis of agricultural policies.In: IIFET 2000: Microbehavior and Macro-results. Proceedings. International Institute of Fisheries Economics and Trade,Oregon State University, Corvallis, Oregon, USA.

Benvenuti, B., 1961. Farming in cultural change. Van Gorcum, Assen.Berdegue, J., Escobar, G., 1990. Efectos de la metodologıa de tipificacion en la

investigacion de sistemas de produccion. In: Escobar, G., Berdegue, J. (Eds.), Ti-pificacion de Sistemas de Produccion. RIMISP, Santiago de Chile, pp. 251–265.

Berger, T., Schreinemachers, P., 2006. Creating agents and landscapes for MAS from

random samples. Ecology and Society 11, 19.Berkhout, P., Bruchem, C. van, 2005. Agricultural economic report 2005 of the

Netherlands: summary. In: LEI (Ed.), Den Haag.Bonabeau, E., 2002. Agent-based modeling: methods and techniques for simulating

human systems. Proceedings of National Academy of Sciences 99, 7280–7287.Bousquet, F., Le Page, C., Bakam, I., Takforyan, A., 2001. Multiagent simulations of

hunting wild meat in a village in eastern Cameroon. Ecological Modelling 138,331–346.

Bowler, I., 1992. ‘Sustainable agriculture’ as an alternative path of farm businessdevelopment.In: Bowler, I., Bryant, C.R.,Nellis, M.D. (Eds.), Contemporary RuralSystems in Transitions. Redwood Press, Melksham.

Brown, D.G., 2005. Agent-based models. In: Geist, H.J. (Ed.), Our Earth’s ChangingLand: An Encyclopedia of Land-Use and Land-Cover Change. Greenwood Press.

Bryant, C.R., Johnston, T.R.R., 1992. Agriculture in the City’s Countryside. BelhavenPress, London.

Burton, R.J.F., Wilson, G.A., 2006. Injecting social psychology theory into concep-tualisations of agricultural agency: towards a post-productivist farmer self-identity? Journal of Rural Studies 22, 95–115.

Busck, A.G., 2002. Farmers’ landscape decisions: relationships between farmers’

values and landscape practices. Sociologia Ruralis 42, 233–249.Castella, J.-C., Boissau, S., Trung, T.N., Quang, D.D., 2005. Agrarian transition and

lowland-upland interactions in mountain areas in northern Vietnam: applica-tion of a multi-agent simulation model. Agricultural Systems 86, 312–332.

Commandeur, M.A.M., 2005. Styles of pig farming and family labour in the Nether-lands. Journal of Comparative Family Studies 36, 391–398.

Crawford, T.W., Messina, J.P., Manson, S.M., David, O.S., 2005. Complexity science,complex systems,and land-useresearch.Environment and Planning B-Planning& Design 32, 792–798.

de Lauwere, C.C., 2005. The role of agricultural entrepreneurship in Dutch agri-culture of today. Agricultural Economics 33, 229–238.

Escobar, G., Berdegue, J., 1990. Conceptos y metodologıas para la tipificacion d esistemas de finca: la experiencia de RIMISP. In: Escobar, G., Berdegue, J.(Eds.), Tipificacion de Sistemas de Produccion. RIMISP, Santiago de Chile, pp.13–43.

Evans, N.J.,Ilbery, B.W.,1989. A conceptual frameworkfor investigating farm-basedaccommodation and tourism in Britain. Journal of Rural Studies 5, 257–266.

Evans, T.P., Kelley, H., 2004. Multi-scale analysis of a household level agent-based

model of landcover change. Journal of Environmental Management 72, 57–72.




Fairweather, J.R., Keating, N.C., 1994. Goals and management styles of New Zealandfarmers. Agricultural Systems 44, 181–200.

Ferrand, N., 1996. Modeling and supporting multi-actor spatial planning usingmulti-agents systems. In: Third International Conference Integrating GIS andEnvironmental Modeling, Santa Fe.

Fish, R., Seymour, S., Watkins, C., 2003. Conserving English landscapes: landmanagers and agri-environmental policy. Environment and Planning A 35,19–41.

Fohrer, N., Moller, D., Steiner, N., 2002. An interdisciplinary modelling approach toevaluate theeffectsof land usechange.Physicsand Chemistry ofthe Earth,Parts

A/B/C 27, 655–662.Gasson, R., 1973. Goals and values of farmers. Journal of Agricultural Economics 24,

521–542.Geertman, S., Stillwell, J., 2003. Planning support systems in practice. In: Advances

in Spatial Science, Springer Verlag, Berlin, p. 578.Gimblett, H.R., 2002. Integrating geographic information systems and agent-based

technologies for modeling and simulating social and ecological phenomena. In:Gimblett, H.R. (Ed.), Integrating Geographic Information Systems and Agent-based Modeling Techniques for Simulating Social and Ecological Processes.Oxford University Press, New York, pp. 1–20.

Girard, N., 2006. Categoriser les practiques d’agriculteurs pour reformuler unprobleme en partenariat: une proposition methodologique. Cahiers Agricul-tures 15, 261–272.

Graveland,C., van Rheenen,T., Brouwer,F., 2004.Countrystudy for the Netherlands.In: BiofACT, LEI, The Hague.

Guillaumin, A., Bousquet, D., Villaret, A., 2004. Multifonctionnalite de l’agriculture:demandes locales et attitudes des agriculteurs. In: Laurent, C., Remy, J. (Eds.),Les Cahiers de la Multifonctionnalite. INRA-CEMAGREF-CIRAD.

Happe, K., Kellermann, K., Balmann, A., 2006. Agent-based analysis of agricultural

policies: an illustration of the agricultural policy simulator AgriPoliS, its adapta-tion and behavior. Ecology and Society 11, 49., http://www.ecologyandsociety.org/vol11/iss1/art49/.

Huigen, M.G.A., 2004. First principles of the MameLuke multi-actor modellingframeworkfor land use change, illustrated with a Philippine case study.Journalof Environmental Management 72, 5–21.

Ilbery, B.W., 1978. Agricultural decision-making: a behavioural perspective. Pro-gress in Human Geography 2, 448–466.

Janssen, M.A., Ostrom, E., 2006. Empirically based, agent-based models. Ecology andSociety 11, 37–49.

Jepsen, M.R., Leisz, S., Rasmussen, K., Jakobsen, J., Møller-Jensen, L., Christiansen, L.,2006. Agent-based modelling of shifting cultivation field patterns, Vietnam.International Journal of Geographical Information Science 20, 1067–1085.

Jollivet, M., 1965. D’une methode typologique pour l’etude des societes rurales.Revue Francaise de Sociologie VI, 33–54.

Jongeneel, R., Polman, N.N., Slangen, L., 2005. Multifunctional alternatives foragriculture in changing landscapes: understanding multifunctionality in theWinterswijk region-case. In: Multifunctionality of Landscapes, Germany.

Jongeneel, R.A., Polman, N.B.P., Slangen, L.H.G., 2008. Why are Dutch farmers g oingmultifunctional? Land Use Policy 25, 81–94.Kelly, C.E., Ilbery, B.W., 1995. Defining and examining rural diversification – a

framework for analysis. Tijdschrift Voor Economische En Sociale Geografie 86,177–185.

Knickel, K., Renting, H., 2000. Methodological and conceptual issues in the study of multifunctionality and rural development. Sociologia Ruralis 40, 512–528.

Knowler, D., Bradshaw, B., 2007. Farmers’ adoption of conservation agriculture: areview and synthesis of recent research. Food Policy 32, 25–48.

Kobrich, C., Rehman, T., Khan, M., 2003. Typification of farming systems for con-structing representative farm models: two illustrations of the application of multi-variate analyses in Chileand Pakistan. Agricultural Systems 76, 141–157.

Kristensen, S.P., 2003. Multivariate analysis of landscape changes and farm char-acteristicsin a study area in central Jutland,Denmark.Ecological Modelling 168,303–318.

Lambin, E.F., Geist, H.J., 2001. Global land-use and cover change: What have welearned so far? Global Change Newsletter No. 46, 27–30.

Landais, E.,1998. Modelling farm diversity:new approaches to typology building inFrance. Agricultural Systems 58, 505–527.

Legendre, P., Legendre, L., 1998. Numerical Ecology. Elsevier, Amsterdam.Lesschen, J.P., Verburg, P.H., Staal, S.J., 2005. Statistical methods for analysing the

spatial dimension of changes in land use and farming systems. In: LUCC, IHDP,IGBP (Eds.), LUCC Report Series 7.

Ligtenberg, A., Wachowicz, M., Bregt, A.K., Beulens, A., Kettenis, D.L., 2004. A designand application of a multi-agent system for simulation of multi-actor spatialplanning. Journal of Environmental Management 72, 43–55.

Luijten, J.C., 2003. A systematic method for generating land use patterns usingstochastic rules and basic landscape characteristics: results for a Colombianhillside watershed. Agriculture, Ecosystems & Environment 95, 427–441.

Luttik, J., van der Ploeg, B., 2004. Functions of agriculture in urban society in theNetherlands. In: Brouwer, F. (Ed.), Sustaining Agriculture and the Rural Envir-onment: Governance, Policy and Multifunctionality. Edward Elgar Publishing,Cheltenham, UK, pp. 204–222.

Mastboom, J.M.J., 1996. Protoindustrialization and agriculture in the eastern Neth-erlands. Social Science History 20, 235–258.

Matthews, R., 2006. The People and Landscape Model (PALM): towards full inte-gration of human decision-making and biophysical simulation models. Ecolo-gical Modelling 194, 329–343.

Matthews, R., Gilbert, N., Roach, A., Polhill, J., Gotts, N., 2007. Agent-based land-usemodels: a review of applications. Landscape Ecology 22, 1447–1459.

McIntosh, B.S., Seaton, R.A.F., Jeffrey, P., 2007. Tools to think with? Towards under-standing the use of computer-based support tools in policy relevant research.Environmental Modelling & Software 22, 640–648.

McKinney, J.C., 1950. The role of constructive typology in scientific sociologicalanalysis. Social Forces 28, 235–240.

Meert, H., Van Huylenbroeck, G., Vernimmen, T., Bourgeois, M., van Hecke, E., 2005.Farm household survival strategies and diversification on marginal farms.

Journal o f Rural Studies 2 1, 81–97.

Monticino, M., Acevedo, M., Callicott, B., Cogdill, T., Lindquist, C., 2007. Coupledhuman and natural systems: a multi-agent-based approach. EnvironmentalModelling & Software 22, 656–663.

Morris, C., Potter, C., 1995. Recruiting the new conservationists: farmers’adoption of agri-environmental schemes in the UK. Journal of Rural Studies11, 51–63.

Oerlemans, N., van Well, E., Guldemond, A., 2004. Agrarische natuurverenigingenaan de slag: een tweede verkenning naar de rol van agrarische natuutver-enigingen in natuurbeheer. In: Centrum voor Landbouw en Milieu, Culem-borg, p. 30.

Ondersteijn, C.J.M., Giesen, G.W.J., Huirne, R.B.M., 2003. Identification of farmercharacteristics and farm strategies explaining changes in environmental man-agement and environmental and economic performance of dairy farms. Agri-cultural Systems 78, 31–55.

Parker, D.C., Berger, T., 2002. Synthesis and discussion. In: Parker, D.C., Berger, T.,Manson, S.M. (Eds.), Agent-based models of land-use and land-coverchange: report and review of an international workshop, Irvine, California,pp. 79–88.

Parker, D.C., Manson, S.M., Janssen, M.A., Hoffmann, M.J., Deadman, P., 2003. Multi-

agent systems for the simulation of land-use and land-cover change: a review.Annals of the Association of American Geographers 93, 314–337.

Perrot, C., Landais, E., 1993a. Comment modeliser la diversite des exploitationsagricoles? Les Cahiers de la Recherche Developpement 24–40.

Perrot, C., Landais, E., 1993b. Exploitations agricoles: pourquoi poursuivre larecherche sur les methodes typologiques? Les Cahiers de la Recherche Devel-oppement 13–23.

Primdahl, J., 1999. Agricultural landscapes as places of production and for living inowner’s versus producer’s decision making and the implications for planning.Landscape and Urban Planning 46, 143–150.

Provinciale Statenvan Gelderland,2005a. Streekplan Gelderland 2005:kansen voorde regio’s, p. 161.

Provinciale Staten van Gelderland, 2005b. Grond voor verandering: reconstructieAchterhoek en Liemers, p. 134.

Rindfuss, R.R., Walsh, S.J., Turner II, B.L., Fox, J., Mishra, V., 2004. Developing ascience of land change: challenges and methodological issues. Proceedings of National Academy of Sciences 101, 13976–13981.

Robinson, D.T., Brown, D.G.,Parker, D.C.,Schreinemachers, P., Janssen,M.A., Huigen,

M., Wittmer, H., Gotts, N., Promburom, P., Irwin, E., Berger, T., Gatzweiler, F.,Barnaud, C., 2007. Comparison of empirical methods for building agent-basedmodels in land use science. Journal of Land Use Science 2, 31–55.

Rokeach, M., 1968. Theory of organization and change within value-attitudesystems. Journal of Social Issues 24, 13–33.

Sawyer, R.K., 2003. Artificial societies: multiagent systems and the micro-macrolink in sociological theory. Sociological Methods Research 31, 325–363.

Schmitzberger, I., Wrbka, T., Steurer, B., Aschenbrenner, G., Peterseil, J., Zechme-ister, H.G., 2005. How farming styles influence biodiversity maintenance inAustrian agricultural landscapes. Agriculture, Ecosystems & Environment108, 274–290.

Sharpley, R., Vass, A., 2006. Tourism, farming and diversification: an attitudinalstudy. Tourism Management 27, 1040–1052.

Siebert, Rosemarie, Toogood, Mark, Knierim, Andrea, 2006. Factors affecting Eur-opeanfarmers’participation in biodiversitypolicies. Sociologia Ruralis 46, 318–340.

Simon, H.A., 1955. A behavioral model of rational choice. The Quarterly Journal of Economics 69, 99–118.

Soares-Filho, B.S., Coutinho Cerqueira, G., Lopes Pennachin, C., 2002. DINAMICA – a

stochastic cellular automata model designed to simulate the landscapedynamics in an Amazonian colonization frontier. Ecological Modelling 154,217–235.

Speybroeck, N., Berkvens, D., Mfoukou-Ntsakala, A., Aerts, M., Hens, N., Van Huy-lenbroeck, G., Thys, E., 2004. Classification trees versus multinomial models inthe analysis of urbanfarming systems in Central Africa. Agricultural Systems 80,133–149.

Uran, O., Janssen, R., 2003. Why are spatial decision support systems not used?Some experiences from the Netherlands. Computers, Environment and UrbanSystems 27, 511–526.

van der Ploeg, J.D., 1994. Styles of farming: an introductory note on concepts andmethodology. In: van der Ploeg, J.D.,Long, A. (Eds.), Practice and Perspectives of Endogenous Rural Development. Van Gorcum, Assen, pp. 7–31.

van der Ploeg, J.D., 2003. The virtual farmer. Royal van Gorcum, Assen.van der Veen, A., Otter, H.S., 2001. Land use changes in regional economic theory.

Environmental Modeling and Assessment 6, 145–150.Van Doorn, A., Bakker, M., 2007. The destination of arable land in a marginal

agricultural landscape in South Portugal: an exploration of land use changedeterminants. Landscape Ecology 22, 1073–1087.


http://www.ecologyandsociety.org/vol11/iss1/art49/






van Horne, P., Prins, H., 2002. Development of dairy farming in the Netherlands inthe period 1960–2000. In: LEI, The Hague, p. 24.

Vanclay, F., Howden, P., Mesiti, L., Glyde, S., 2006. The social and intellectualconstruction of farming styles: testing Dutch ideas in Australian agriculture.Sociologia Ruralis 46, 61–82.

Verburg, P.H., 2006. Simulating feedbacks in land use and land cover changemodels. Landscape Ecology V21, 1171–1183.

Verburg, P.H.,Soepboer,W., Veldkamp,A., Limpiada,R., Espaldon,V., Mastura, S.S.A.,2002. Modeling the spatial dynamics of regional land use: The CLUE-S model.Environmental Management 30, 391–405.

Wildenbeest, G., 1989. Recent farmers’ protest in a Dutch municipality: the legacyof the past. In: Boissevain, J., Verrips, J. (Eds.), Dutch Dilemmas: Antropologistslook at the Netherlands. Van Gorcum, Assen, pp. 70–88.

Willock, J., Deary, I.J., Edwards-Jones, G., Gibson, G.J., McGregor, M.J., Sutherland, A.,Dent, J.B., Morgan, O., Grieve, R., 1999a. The role of attitudes and objectives infarmer decision making: business and environmentally-oriented behaviour inScotland. Journal of Agricultural Economics 50, 286–303.

Willock, J., Deary, I.J.,McGregor, M.M.,Sutherland, A., Edwards-Jones,G., Morgan, O.,Dent, B., Grieve, R., Gibson, G., Austin, E., 1999b. Farmers’ attitudes, objectives,behaviors, and personality traits: the Edinburgh study of decision making onfarms. Journal of Vocational Behavior 54, 5–36.

Wilson, G.A., 2007. Multifunctional Agriculture: A Transition Theory Perspective.Cromwell Press, Trowbridge.

Wilson, G.A., Hart, K., 2000. Financial imperative or conservation concern? EUfarmers’ motivations for participation in voluntary agri-environmentalschemes. Environment and Planning A 32, 2161–2185.


valbuena et al 2008 aee

Documents